기계학습을 이용한 파레토 프런티어의 생성

윤예분,정나영,윤민,Yun, Yeboon,Jung, Nayoung,Yoon, Min 한국데이터정보과학회 2013 한국데이터정보과학회지 Vol.24 No.3

진화 알고리즘 계산 지능을 이용한 예측 방법이 다목적 최적화 문제에서 많이 이용되고 있고, 이러한 방법들은 많은 근사 파레토 최적해들을 좀 더 정확하게 생성하기 위해서 개선되고 있다. 본 논문은 다목적 최적화 문제에서 서포트 벡터기계를 이용하여 근사 파레토 프런티어를 찾는 방법을 제안한다. 또한 제안된 방법과 진화 알고리즘을 결합한 것이 파레토 프런티어를 더 잘 근사시킨다는 것과 두 개혹은 세 개의 목적함수를 가진 의사결정은 제안된 방법으로 파레토 프런티어를 시각화한 것에 근거하여 더 쉽게 수행된다는 것을 보인다. 마지막으로 몇 개의 수치예제를 통해 제안된 방법의 효율성에 대해 보일 것이다. Evolutionary algorithms have been applied to multi-objective optimization problems by approximation methods using computational intelligence. Those methods have been improved gradually in order to generate more exactly many approximate Pareto optimal solutions. The paper introduces a new method using support vector machine to find an approximate Pareto frontier in multi-objective optimization problems. Moreover, this paper applies an evolutionary algorithm to the proposed method in order to generate more exactly approximate Pareto frontiers. Then a decision making with two or three objective functions can be easily performed on the basis of visualized Pareto frontiers by the proposed method. Finally, a few examples will be demonstrated for the effectiveness of the proposed method.

다목적 유전자 알고리즘에 있어서 적합도 평가방법과 대화형 의사결정법의 제안

윤예분,박동준,윤민 한국산업경영시스템학회 2022 한국산업경영시스템학회지 Vol.45 No.4

Most of real-world decision-making processes are used to optimize problems with many objectives of conflicting. Since the betterment of some objectives requires the sacrifice of other objectives, different objectives may not be optimized simultaneously. Consequently, Pareto solution can be considered as candidates of a solution with respect to a multi-objective optimization (MOP). Such problem involves two main procedures: finding Pareto solutions and choosing one solution among them. So-called multi-ob- jective genetic algorithms have been proved to be effective for finding many Pareto solutions. In this study, we suggest a fitness evaluation method based on the achievement level up to the target value to improve the solution search performance by the multi-objective genetic algorithm. Using numerical examples and benchmark problems, we compare the proposed method, which considers the achievement level, with conventional Pareto ranking methods. Based on the comparison, it is verified that the proposed method can generate a highly convergent and diverse solution set. Most of the existing multi-objective genetic algorithms mainly focus on finding solutions, however the ultimate aim of MOP is not to find the entire set of Pareto solutions, but to choose one solution among many obtained solutions. We further propose an interactive decision-making process based on a visualized trade-off analysis that incorporates the satisfaction of the decision maker. The findings of the study will serve as a reference to build a multi-objective decision-making support system.

서포트 벡터 기계에서 메타 학습법에 대한 연구

윤예분(Yeboon Yun),윤민(Min Yoon) 한국자료분석학회 2021 Journal of the Korean Data Analysis Society Vol.23 No.1

순차적 근사 최적화는 계산 비용이 많이 드는 함수들을 갖는 많은 공학 문제를 해결하기 위하여 적용되는 방법이다. 메타 모형(근사함수 형식)은 몇개의 표본을 기반으로 구성되며 한 번에몇 개의 표본을 추가하여 순차적으로 개선된다. 서포트 벡터 회귀(support vector regression) 및 방사형 기저 함수 네트워크(radial basis function networks)와 같은 계산 지능이 메타 모델화에 효과 적으로 적용될 수 있음이 알려져 왔다. 모형의 성능은 가우시안 함수의 폭 및 정규화 매개 변수와 같은 초모수(hyper-parameter)들의 영향을 받는다. 적합하거나 적절한 초모수들을 추정하는 가장 널리 사용되는 방법들 중에서 대표적인 방법은 교차 타당성 검정이다. 그러나 교차 타당성 검정은 일반적으로 시간이 많이 소요되며 표본점들의 수가 적은 순차적 근사 최적화에는 적용되지 않을 수 있다. 본 연구에서는 가우시안 함수의 폭을 선택하는 부담을 줄이고 가능한 적은 수의 표본들로 좋은 근사 모델을 생성하기 위하여 배깅(bagging)과 부스팅(boosting)을 사용하는 순차적 메타 학습 방법을 제안한다. 또한 제안된 방법의 효율성을 수치 실험을 통하여 일반화 능력과 모수의 조절이라는 관점에서 조사하였다. Sequential approximate optimization is applied for solving many engineering problems with computationally expensive functions. Meta model (approximated function form) is constructed based on some samples and is improved in sequence by adding a few samples at a time. It has been observed that computational intelligence such as support vector regression(SVR) and radial basis function networks(RBFN) can be effectively applied to meta-modeling. The performance is affected by hyper-parameters, for example, Gaussian width and a regularization parameter. One of the most popular methods for estimating suitable/adequate hyper-parameters is the cross validation(CV) test. However, CV test is usually time consuming, and, may not be applicable for sequential approximation optimization with a small number of sample points. In this research, we suggest a sequential meta-learning method using bagging and boosting in order to reduce the burden on choosing the width of Gaussian function as well as to generate a good approximate model with as a small number of samples as possible. In addition, the effectiveness of the proposed method is investigated in terms of generalization ability and control of parameter through numerical experiments.

품질경쟁력 우수기업의 특성분석

박동준,윤예분,강인선,유은재,김호균,윤민 한국산업경영시스템학회 2019 한국산업경영시스템학회지 Vol.42 No.3

Quality management has become an pervasive philosophy in most sectors of business. Specific movements such as statistical quality control, quality circle, total quality management, and quality management system have become embedded in business organizations. Only the companies with competitive edge can survive in the competition in global market. KSA(Korean Standards Association) established in 1962 has launched all kinds of quality education, quality standard certification service for business, and KNQA(Korean National Quality Award) system. This article considers quality competitiveness excellent company award among KNQA. We performed a statistical analysis of audit data for quality competitiveness excellent company for three years, from 2015 to 2017. By using ANOVA and two sample t-tests, the average scores of 13 evaluation fields were significantly different depending on company size and type. We proposed ways to improve the current hall of fame system. We discovered that the average scores of 13 evaluation fields in the audit data according to years and hall of fame status were not significantly different. We also showed linear relationships among 13 evaluation fields by correlation analysis and obtained an estimated linear regression equation : Business Performance, which is a comprehensive index, as a dependent variable was significantly related to Customer Focus and Product Liability as regressor variables among 13 evaluation fields by regression analysis.

기계학습 방법을 이용한 기업부도의 예측

박동준,윤예분,윤민,Park, Dong-Joon,Yun, Ye-Boon,Yoon, Min 한국데이터정보과학회 2012 한국데이터정보과학회지 Vol.23 No.3

The analysis and management of business failure has been recognized to be important in the area of financial management in the evaluation of firms' performance and the assessment of their viability. To this end, effective failure-prediction models are needed. This paper describes a new approach to prediction of business failure using the total margin algorithm which is a kind of support vector machine. It will be shown that the proposed method can evaluate the risk of failure better than existing methods through some real data. 기업도산에 대한 분석과 관리는 기업의 성과와 성장능력을 평가하는 재무관리 분야에서 중요하게 인식되어 왔다. 결국, 기업도산 예측에 대한 효과적인 모형이 필요하게 된다. 본 논문은 서포트 벡터 기계의 한 종류인 토탈 여유도 알고리즘을 이용하여 기업도산 예측을 위하여 새로운 접근 방법을 서술한다. 몇 개의 실제 자료를 통하여 제안한 방법들이 도산 위험의 평가에서 기존의 방법들보다 개선됨을 확인할 수 있었다.

기계 학습을 이용한 고객 분류 성능 비교

엄정길,서한손,윤예분,윤민 한국자료분석학회 2012 Journal of the Korean Data Analysis Society Vol.14 No.5

As a tools of pattern classification problem, machine learning has been globally studied. Support vector machines, as a kind of machine learning, is the pattern classification problem with two class sets, recently. This idea is to find a maximal margin separating hyperplane which gives the largest separation between then classes in a high dimensional feature space. Linear classifiers then are optimized to give the maximal margin separation between the classes. The task is performed by solving some type of mathematical programming such as quadratic programming or linear programming. In this paper, compared the customer classification performance using logistic regression and support vector machine with customer credit data. The misclassification rate of logistic regression and support vector machine in testing data is 1.89% and 1.73% respectively. Consequently, the misclassification rate of support vector machine is slightly superior to logistic regression. 패턴분류의 한 도구로서 기계학습은 광범위하게 연구되고 있다. 최근에 기계학습 방법 중에서 서포트 벡터 기계(support vector machines)는 이진 패턴 분류문제에서 고차원의 특징공간에서 두 집합들 사이에 가장 큰 분리를 제공하는 최대 여유도(margin)를 가지는 분리 초평면을 찾는 것이다. 특히 선형 분류기들은 두 집합들 사이를 최대 여유도 분리를 얻기 위하여 최적화한다. 이러한 작업들은 이차계획문제(quadratic programming)나 일차계획문제(linear programming)와 같은 수리계획법의 어떤 형태를 풀어내는 것으로 얻어진다. 본 논문에서는 신용 평가 자료를 이용하여 고객 분류 문제에 있어서 전통적인 분류 방법인 로지스틱 회귀분석과 서포트 벡터 기계를 이용하여 분류 성능을 비교하였다. 테스트 집합에서 로지스틱 회귀분석과 서포트 벡터 기계의 오분류율은 각각 1.89%와 1.73%로 얻어졌다. 결국 서포트 벡터 기계의 오분류율은 로지스틱 회귀분석보다 다소 개선됨을 알 수 있었다.

품질경쟁력 우수기업 평가지표의 확인적 요인분석

박동준(Dong Joon Park),윤예분(Yeboon Yun),윤민(Min Yoon) 한국산업경영시스템학회 2020 한국산업경영시스템학회지 Vol.43 No.3

Companies struggle to make their best products with high quality and service at a competitive price in global markets. However, customer needs and requirements keep changing with a variety of situations. Companies that face the changes can not stay the same and make an effort to adapt themselves to new circumstances. They would probably review the overall management system that is currently implementing to improve management efficiency. Among other things, quality might be considered to be a crucial element if they are manufacturing industries to be sustained in global markets. KSA (Korean Standards Association) is a government-affiliated organization under the Ministry of Trade, Infrastructure, and Energy. It is a Korean standards provider for quality and service industry. KSA confers national commendations for organizations, quality circles, artisans, QCEC (Quality Competitive Excellent Company), and the most honorable KNQA (Korean National Quality Award) every year. KSA established KNQA on the basis of Malcom Baldrige National Quality Award, Deming Prize, and European Quality Award. Research on quality awards shows that there are many similarities in the framework. Although KSA summarizes two factors for 13 evaluation indicators in the quality competitive excellent model of QCEC, the categorization is ambiguous to explain them according to earlier studies. We performed a deep analysis of foreign quality awards and background for KNQA and QCEC. We conducted a content analysis of KNQA and QCEC and matched evaluation items that were closely related. We proposed a quality competitiveness model with three factors, Technology, System, and Tools, summarizing 13 evaluation indicators in QCEC. Based on audit data for six years from 2012 to 2017 we carried out a confirmatory factor analysis for the proposed model by examining the model validity and fitness.